feat(llm): return chat completion token ids via nvext

[AmeenP]

Summary

This is the Dynamo-side token-in/nvext contract needed by PrimeIntellect-ai/renderers#11, split out from the larger draft RL/admin PR.

• adds nvext.extra_fields: ["completion_token_ids"] for chat completions
• accumulates generated token IDs across chat completion chunks and returns them under response.nvext.completion_token_ids on the final chunk/non-streaming aggregate response
• rejects nvext.extra_fields: ["completion_token_ids"] when n > 1 until Dynamo has an indexed per-choice shape
• accepts rl-sdk compatible root stop_token_ids and plumbs them into token stop conditions
• accepts renderer-compatible top-level cache_salt, supports canonical nvext.cache_salt, and forwards the salt through backend extra_args["nvext"]
• supports nvext.extra_fields: ["engine_data"] by returning generated completion_token_ids and available completion_logprobs under response.nvext.engine_data
• documents that nvext.token_data skips tokenization independently of backend_instance_id

Intentionally not included: /v1/rl/*, LoRA admin, pause/resume, weight update, or worker RL dispatch changes from #9382.

Validation

• cargo fmt --check
• git diff --check
• python -m py_compile components/src/dynamo/vllm/handlers.py

Attempted targeted dynamo-llm Rust tests for test_cache_salt, test_completion_token_ids_rejects_multiple_choices, and test_engine_data_accumulates_completion_token_ids_and_logprobs, but this macOS checkout is blocked by existing Linux-only dynamo-llm compile paths (dynamo_memory::numa, DiskStorage, fallocate, O_DIRECT).

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

👋 Hi AmeenP! Thank you for contributing to ai-dynamo/dynamo.

Just a reminder: The NVIDIA Test Github Validation CI runs an essential subset of the testing framework to quickly catch errors.Your PR reviewers may elect to test the changes comprehensively before approving your changes.

🚀

[coderabbitai]

Walkthrough

This PR extends the nvext response protocol with a new completion_token_ids field that accumulates backend-generated token IDs across streaming chunks and emits the concatenated list on the final response chunk, with corresponding schema, streaming logic, and documentation updates.

Changes

Completion Token IDs Support

Layer / File(s)	Summary
NvExt Protocol Extension lib/llm/src/protocols/openai/nvext.rs	NvExtResponse and NvExtResponseFieldSelection structs gain completion_token_ids fields; field selection logic parses "completion_token_ids" from extra_fields; documentation and unit tests updated to reflect the new optional field.
DeltaGenerator Streaming Implementation lib/llm/src/protocols/openai/chat_completions/delta.rs	DeltaGenerator adds internal state to accumulate token IDs across chunks; choice_from_postprocessor conditionally appends backend token IDs when the field is enabled; nvext response construction injects the accumulated list into the payload on the final chunk only; unit test verifies correct chunk-level emission.
User-Facing Documentation docs/components/frontend/nvext.md	Field reference table clarified and extra_fields support expanded; response extensions table documents the new completion_token_ids field and its final-chunk emission semantics.

---

🎯 3 (Moderate) | ⏱️ ~22 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'feat(llm): return chat completion token ids via nvext' clearly and specifically summarizes the main change—adding support for returning completion token IDs through the nvext response field.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description check	✅ Passed	The pull request description is comprehensive and well-structured, covering the purpose, implementation details, validation steps, and intentional exclusions.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

Tip

💬 Introducing Slack Agent: The best way for teams to turn conversations into code.

Slack Agent is built on CodeRabbit's deep understanding of your code, so your team can collaborate across the entire SDLC without losing context.

• Generate code and open pull requests
• Plan features and break down work
• Investigate incidents and troubleshoot customer tickets together
• Automate recurring tasks and respond to alerts with triggers
• Summarize progress and report instantly

Built for teams:

• Shared memory across your entire org—no repeating context
• Per-thread sandboxes to safely plan and execute work
• Governance built-in—scoped access, auditability, and budget controls

One agent for your entire SDLC. Right inside Slack.

👉 Get started

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}